Question about Cyrus, ext3, and Linux kernel 2.4.18 bug

[Michael Sims]

Quoting Henrique de Moraes Holschuh <hmh at debian.org>:

> On Fri, 11 Apr 2003, Michael Sims wrote:
> > Is anyone out there running Cyrus on Linux kernel 2.4.18, using ext3 with
> > "data=journal" on the Cyrus spool volume?  If so, have you had any
> problems?
> 
> That will kill your kernel (unless redhat did a major ext3 bughunt on their
> 2.4.18). Don't do it. Switch to 2.4.21pre, or try using xfs.
[...]
> The kernel syslogs errors, and will hang any accesses to the affected
> filesystem (unkillable D state processes).  At that point, umount everything
> you can, and hard-reset the machine, because nothing else other than SysRq+B
> or a hard reset will work.  Filesystem corruption is assured.

Wow.  Thanks for the information.  I just unmounted the drive and switched it 
back to ordered mode.  I guess I just got lucky that it hadn't already happened.

The only reason that I thought this might already be fixed is that I saw a 
couple of Red Hat errata kernels that mentioned ext3 bugs.  For example:

https://rhn.redhat.com/errata/RHBA-2002-085.html

"In the 2.4.18-3 kernel, a timing change exposed a previously-unseen
race condition in ext3 which could occasionally cause the kernel
to panic on SMP systems. The 2.4.18-4 kernel corrects this problem."

Also:

https://rhn.redhat.com/errata/RHBA-2002-292.html

"A bug in the ext3 file system in the previous errata kernel for Red Hat
Linux 7.1, 7.1K, 7.2, 7.3, and 8.0 has been discovered. The bug has the
potential to cause data loss if the file system is used in the non-default
'full data journaling' mode."

Of course this last one goes on to say that the bug "can only result in the 
loss of recent writes when a file system is unmounted" so I don't believe it's 
the same issue.

If performance gets too bad I may just fall back to ext2.  I believe (but could 
be mistaken) that I would have to compile a new kernel to use XFS since Red Hat 
doesn't support it by default.

At any rate, thanks for the information, you probably just saved my job. :)

___________________________________________
Michael Sims
Project Analyst - Information Technology
Crye-Leike Realtors
Office: (901)758-5648  Pager: (901)769-3722
___________________________________________